Data communication system based on multi-source data synchronization and dynamic error compensation

Abstract

The invention relates to the technical field of data communication, and particularly discloses a data communication system based on multi-source data synchronization and dynamic error compensation, which is used for solving the problem that node state recognition errors and data are inconsistent due to environment disturbance, structure offset and image frame loss of communication equipment. The hardware equipment layer comprises a multi-mode sensing terminal, auxiliary imaging equipment, a laser scanning device and an environment monitoring sensor, the multi-mode sensing terminal, the auxiliary imaging equipment and the environment monitoring sensor are all connected with a data acquisition card, the hardware equipment layer is connected with the data acquisition layer, the data acquisition layer is connected with the software system layer, and the software system layer is connected with the function application layer; according to the method, an environment compensation model is constructed by using multi-angle static modeling and disturbance parameters, time-space consistency fusion of image data is realized in combination with optical flow information and I MU attitude data, communication structure state recognition and error compensation are realized through an auto-encoder network and dynamic correction on a Spark platform, and data stability is improved.

Description

Technical Field

[0001] This invention relates to the field of data communication technology, and more specifically, to a data communication system based on multi-source data synchronization and dynamic error compensation. Background Technology

[0002] With the rapid development of the Industrial Internet, intelligent manufacturing, and edge computing, the types of data generated by field devices are becoming increasingly diverse, involving multimodal data sources such as image streams, sensor data, structural information, and environmental disturbance parameters. During transmission, these data are constrained by complex environmental factors (such as electromagnetic interference, vibration noise, and changes in lighting), easily leading to errors, frame drops, and signal distortion, affecting the reliability and accuracy of data communication. Existing data communication systems mostly employ fixed modulation schemes and static filtering algorithms, which struggle to cope with dynamic changes in operating environmental conditions. Especially when the communication structure itself has minor variations or fluctuating operating states, there is a lack of effective error perception and compensation mechanisms. Furthermore, deep integration of environmental parameters and structural change characteristics is not fully utilized; reliance on end-to-end error control strategies results in limited processing accuracy and insufficient robustness, failing to meet the synchronization and reliability requirements of industrial communication scenarios. Therefore, there is an urgent need to propose a communication system for multi-source data communication that combines static structural modeling and dynamic data compensation mechanisms to identify communication errors caused by environmental disturbances or structural offsets in real time. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, this invention provides a data communication system based on multi-source data synchronization and dynamic error compensation. This system can combine static structural modeling and dynamic data compensation mechanisms to identify communication errors caused by environmental disturbances or structural offsets in real time. Furthermore, it enhances the system's anti-interference capability and adaptive adjustment capability through fusion computing and model training, thereby improving the stability and intelligence of data transmission.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A data communication system based on multi-source data synchronization and dynamic error compensation includes a hardware device layer, which comprises a multimodal sensing terminal, an auxiliary imaging acquisition unit, a laser scanning device, and an environmental monitoring module. The multimodal sensing terminal, auxiliary imaging unit, and environmental monitoring module are all connected to data acquisition cards. The hardware device layer connects to a data acquisition layer, which in turn connects to a software system layer, which in turn connects to a functional application layer. The hardware device layer is used to acquire structural point cloud data and environmental information of communication equipment components in both static and dynamic states. The data acquisition layer extracts errors during the operation of the dynamic communication equipment by analyzing the static point cloud information. The software system layer uses a static benchmark model and environmental monitoring data to dynamically correct and compensate for the errors, and corrects the positioning coordinates of the communication nodes through multi-source fusion and filtering techniques. The functional application layer constructs a three-dimensional model of the communication components and a digital twin system based on the corrected data, enabling remote monitoring, communication flow analysis, and anomaly early warning.

[0006] As a further embodiment of the present invention, the multimodal sensing terminal includes a sensing head component with a reflective target on it. An imaging lens is provided on one side of the reflective target. The imaging lens is connected to an image sensor, and the image sensor is connected to a signal processing module. The surface of the reflective target is provided with an anti-interference multilayer film structure, which includes, from the inside to the outside:

[0007] First coating: 85nm thick silicon nitride interference film, used for initial reflection equalization;

[0008] The second coating is an aluminum oxide electromagnetic shielding layer with a thickness of 50nm, used to reduce high-frequency noise interference.

[0009] The third coating is a 90nm thick silicon carbide structure film, which improves the mechanical stability and adhesion of the film system.

[0010] Fourth coating: A 55nm thick zinc oxide antistatic film, used to reduce surface charge accumulation;

[0011] The fifth coating is an 85nm thick magnesium fluoride antireflective film used to optimize the suppression performance of incident angle light reflection.

[0012] As a further embodiment of the present invention, in the hardware device layer, the sensing head component senses the connection status, spatial changes, and relative posture information of the communication interface area; the auxiliary imaging acquisition unit captures the reflected mark image for visual reference positioning from a set angle; the laser scanning device supplements the acquisition of the three-dimensional point cloud information of the communication component and identifies the connection status and spatial configuration of the communication component; the environmental monitoring module acquires disturbance data of temperature, humidity, and vibration for use as a dynamic compensation reference; the data acquisition card acquires the raw signal and outputs it to the data acquisition layer through the CAN bus; the communication node structure information includes, but is not limited to, the edge contour, contact curvature, and transition area morphology of the fiber optic interface, RF connector, and cable terminal; the image acquisition direction is based on the normal direction of a communication port, and the viewing angle is set every 2° within a ±30° angle range.

[0013] As a further aspect of this invention, when the communication device is in a static calibration state, the data acquisition layer acquires point cloud and image data of the communication nodes of the static structural components. It uses a calibration template and ICP algorithm to uniformly register data acquired from different angles to the global coordinate system, constructs a static reference model, and stores it in the database. During the operation and testing of the communication device, the data acquisition layer adopts an exposure mechanism with an exposure time of less than 5ms, combined with synchronously triggered pulsed illumination, to instantaneously freeze the dynamic process. The dynamically acquired data is divided into several slices, which are processed in parallel by each Worker node in the Spark platform. Simultaneously, environmental disturbance data, including temperature, humidity, vibration, and illumination parameters, are acquired synchronously to construct a time series for error modeling. By aligning the dynamic data with the static reference model, the dynamic error characteristics of the communication node caused by vibration, thermal expansion, and structural displacement during operation are extracted.

[0014] As a further aspect of the present invention, in the data acquisition layer, the static structural components acquired include, but are not limited to, fiber optic patch cord interfaces, base station transceiver connectors, feeder channel ports, antenna power supply devices, and key connection points of RF amplifier modules.

[0015] As a further aspect of the present invention, the software system layer utilizes the raw data from the data acquisition layer to calculate environmental impact factors based on a preset environmental compensation model, corrects the image preprocessing algorithm and optimizes the filter covariance parameters, dynamically adjusts the correction strategy based on the disturbance values ​​of temperature, humidity, and vibration, calculates the communication structure error based on the deviation between the static model and real-time data, smooths the error trajectory through Kalman filtering, updates the node coordinates in real time, and trains an autoencoder network on the Spark platform through stochastic gradient descent for feature extraction and noise suppression.

[0016] As a further aspect of this invention, in the data acquisition layer, the environmental compensation model, based on simulated communication scene data, constructs a linear weighted model through regression analysis, and corrects the image enhancement parameters according to illumination, temperature, humidity, and vibration disturbances. The exposure time and gain are calculated as follows:

[0017]

[0018]

[0019] Where T and G are the exposure value and gain value, respectively, T0 and G0 are the original exposure value and original gain value, respectively, and W... L As a perturbation weighting factor, the environmental monitoring module collects current illumination, current ambient temperature, current relative humidity, and current vibration acceleration in real time during system operation. During the offline calibration phase of the system, the mapping relationship between several sets of environmental variables and the optimal image exposure value is collected to form a sample set. A weighting model is established for the perturbation variables using the multiple linear regression method. The weighted sum of the differences between the current value and the standard value of each perturbation variable is used as the perturbation weighting factor. The weight coefficients of the differences between the current value and the calibration value of each perturbation variable are obtained through training.

[0020] As a further aspect of the present invention, at the software system layer, on the Spark platform, the autoencoder training process includes:

[0021] Step 1: The master node broadcasts the network size, random number seed τ, and learning rate η to each worker node, and at the same time divides the preprocessed data into shards;

[0022] Step 2: Each Worker independently builds an autoencoder to perform forward propagation, error backpropagation, and parameter updates on the samples;

[0023] Step 3: Each worker uploads its local parameters to the master node, and the master node aggregates, updates, and broadcasts the new model;

[0024] Step 4: The trained model generates a correction factor to correct the coordinates of the communication node data.

[0025] As a further aspect of the present invention, in step 1, each Worker node synchronously receives multi-view image streams and IMU and optical flow data provided by the auxiliary imaging unit to construct a multimodal dataset;

[0026] In step 2, the spatiotemporal consistency mask of the image is calculated by combining optical flow and IMU, occlusion features are predicted and missing regions are compensated, and the fused feature data is input into the autoencoder for training.

[0027] As a further aspect of the present invention, the functional application layer uses the corrected data to construct a three-dimensional model of the communication system, and displays the status changes, error trends and key indicator curves of the communication nodes in real time through a digital twin platform, supporting alarm output, operation traceability and automatic report generation.

[0028] The technical advantages of this invention's data communication system based on multi-source data synchronization and dynamic error compensation are as follows:

[0029] This invention achieves a high degree of integration of data acquisition, synchronization, modeling, and prediction for communication nodes in complex industrial environments by introducing multimodal sensing, dynamic error compensation, environmental disturbance modeling, and autoencoder training mechanisms. It can identify error fluctuations caused by disturbances such as temperature, humidity, vibration, and illumination, as well as micro-offsets in the communication structure in real time. Based on the alignment of static models and dynamic data, a dynamic error field is constructed. Combined with Kalman filtering and linear regression, image exposure and gain parameters are dynamically corrected, effectively improving the clarity, stability, and transmission accuracy of communication images and point clouds. At the same time, the autoencoder is trained in parallel using the Spark platform to complete high-dimensional noise suppression and feature extraction, giving the system stronger anti-interference capabilities, adaptive adjustment capabilities, and online modeling capabilities. Finally, a digital twin communication model is constructed through the functional application layer. On the basis of ensuring data synchronization, link stability, and fault prediction, it realizes high-fidelity data transmission and intelligent operation and maintenance in industrial-grade communication scenarios. Attached Figure Description

[0030] Figure 1 This is a system architecture diagram of the present invention;

[0031] Figure 2 This is a schematic diagram of the anti-interference coating structure of the present invention;

[0032] In the diagram: 51 - first coating, 52 - second coating, 53 - third coating, 54 - fourth coating, 55 - fifth coating. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] like Figure 1As shown, the present invention proposes a data communication system based on multi-source data synchronization and dynamic error compensation, comprising a hardware device layer, which includes a multimodal sensing terminal, an auxiliary imaging acquisition unit, a laser scanning device, and an environmental monitoring module. The multimodal sensing terminal, auxiliary imaging unit, and environmental monitoring module are all connected to data acquisition cards. The hardware device layer is connected to a data acquisition layer, which in turn is connected to a software system layer, which is connected to a functional application layer. The hardware device layer is used to acquire structural point cloud data and environmental information of communication equipment components in static and dynamic states. The data acquisition layer extracts errors during the operation of dynamic communication equipment by analyzing static point cloud information. The software system layer uses a static benchmark model and environmental monitoring data to dynamically correct and compensate for errors, and corrects the positioning coordinates of communication nodes through multi-source fusion and filtering techniques. The functional application layer constructs a three-dimensional model of the communication components and a digital twin system based on the corrected data, realizing remote monitoring, communication flow analysis, and anomaly early warning.

[0035] In typical industrial communication systems, such as smart factories, edge data centers, high-precision equipment interconnection platforms, or wireless sensor networking scenarios, communication nodes typically exist in various physical interface forms, including fiber optic interfaces, RF connectors, and cable terminals, and are integrated into dynamically operating equipment (such as motion arms, high-speed rotating components, and equipment susceptible to environmental disturbances). These nodes face several challenges in actual operation:

[0036] (1) Significant environmental disturbances: Environmental factors such as electromagnetic interference, thermal vibration, and humidity fluctuations can easily cause communication link fluctuations, image blurring or point cloud distortion, affecting data synchronization and analysis accuracy.

[0037] (2) Dynamic changes in structural state: The communication interface may undergo slight deformation, displacement, loosening, thermal expansion and other structural shifts during equipment operation, which are difficult to identify through traditional static calibration methods;

[0038] (3) Difficulty in aligning heterogeneous data from multiple sources: The communication system contains heterogeneous information streams such as images, 3D structures, environmental sensors, and IMUs. Traditional acquisition systems have difficulty in achieving time-series alignment and multimodal fusion processing.

[0039] (4) Errors are difficult to compensate dynamically: Existing systems mostly rely on static filtering or fixed modulation methods, which cannot perform intelligent and adaptive real-time error correction in response to environmental and structural disturbances.

[0040] Therefore, the multi-level communication system architecture set in this invention has the following specific capabilities:

[0041] The hardware device layer focuses on data source acquisition and simultaneously constructs a perception scene through structured light, image, and environmental perception.

[0042] The data acquisition layer provides static modeling and dynamic tracking comparison to capture the trends of deformation and error changes;

[0043] The software system layer achieves dynamic error correction through regression models and Kalman filtering;

[0044] The functional application layer constructs digital twins of communication components based on high-quality data to achieve remote operation and maintenance, communication status analysis, and fault early warning.

[0045] It should be noted that, as Figure 2 As shown, the multimodal sensing terminal includes a sensing head component with a reflective target on it. An imaging lens is located on one side of the reflective target, and the imaging lens is connected to an image sensor. The image sensor is connected to a signal processing module. The surface of the reflective target is provided with an anti-interference multilayer film structure, which includes, from the inside out:

[0046] First coating 51: A silicon nitride interference film with a thickness of 85nm, used for initial reflection equalization;

[0047] Second coating 52: Alumina electromagnetic shielding layer with a thickness of 50nm, used to reduce high-frequency noise interference;

[0048] Third coating 53: A silicon carbide structure film with a thickness of 90nm, which improves the mechanical stability and adhesion of the film system;

[0049] Fourth coating 54: A zinc oxide antistatic film with a thickness of 55 nm, used to reduce surface charge accumulation;

[0050] Fifth coating 55: Magnesium fluoride antireflective film with a thickness of 85nm, used to optimize the performance of incident angle light reflection suppression.

[0051] The multimodal sensing terminal of this invention integrates a reflective target onto the sensing head component and constructs a five-layer functional anti-interference film structure on its surface. This enhances image acquisition and positioning accuracy and strengthens the system's robustness in complex communication environments. Specifically, a silicon nitride interferometer is used to adjust the incident light reflection balance and improve image contrast; an alumina shielding layer effectively suppresses high-frequency electromagnetic noise interference to the image sensor; a silicon carbide structural film enhances film adhesion and resistance to mechanical disturbances, adapting to high-frequency vibration and shock in industrial scenarios; a zinc oxide antistatic film reduces the risk of static electricity buildup, preventing image drift and data flicker; and a magnesium fluoride antireflective film further optimizes the reflectivity of incident light at various angles, improving multi-angle imaging clarity. This structure, while ensuring imaging stability, achieves synergistic optimization of environmental interference shielding and reflection compensation, effectively improving the response accuracy and error suppression capability of the image acquisition unit in dynamic communication scenarios. The thicknesses of silicon nitride and magnesium fluoride are set at 85nm each, based on the quarter-wavelength interference principle (λ / 4 refraction matching) at common imaging wavelengths (approximately 450–850nm) in the visible / near-infrared band. This minimizes reflection loss under normal and oblique incidence, improving image clarity and detection sensitivity. The aluminum oxide shielding layer is set at 50nm, forming an effective electromagnetic attenuation layer without affecting optical performance. This suppresses coupling interference from high-frequency radiation (such as antennas and RF modules) from communication systems on the image acquisition device, ensuring that the electrical signal output by the image sensor is not disturbed by clutter. The silicon carbide layer, with a thickness of 90nm, is located in the middle layer. Its high hardness and wear resistance provide physical support, preventing the multilayer film from cracking or peeling due to vibration or thermal expansion and contraction, thus improving the long-term stability of the overall film system in industrial environments. The zinc oxide antistatic layer is set at 55nm, forming a uniform charge dissipation path on the surface of the imaging window, avoiding image shift, color difference, or dead spots caused by static electricity accumulation, and improving image consistency of the system under dry or high-friction conditions.

[0052] It should be noted that in the hardware device layer, the sensor head component senses the connection status, spatial changes, and relative posture information of the communication interface area. The auxiliary imaging acquisition unit captures the reflected mark image for visual reference positioning from a set angle. The laser scanning device supplements the acquisition of the three-dimensional point cloud information of the communication component and identifies the connection status and spatial configuration of the communication component. The environmental monitoring module collects disturbance data of temperature, humidity, and vibration for use as a dynamic compensation reference. The data acquisition card collects the raw signal and outputs it to the data acquisition layer through the CAN bus. The communication node structure information includes, but is not limited to, the edge contour, contact curvature, and transition area morphology of the fiber optic interface, RF connector, and cable terminal. The image acquisition direction is based on the normal direction of a communication port, and the viewing angle is set every 2° within a range of ±30°.

[0053] By setting up a sensor head component in the hardware device layer to perceive the connection status, spatial changes, and relative attitude information of the communication interface area, combined with an auxiliary imaging acquisition unit to capture reflected marker images from multiple preset angles, and with the laser scanning device to supplement the acquisition of three-dimensional point cloud information of the communication components, and with the environmental monitoring module to collect disturbance data such as temperature, humidity, and vibration in real time for dynamic compensation, this system not only achieves high-precision acquisition of key features such as the structural status, spatial deformation, and connection integrity of communication nodes during operation, but also enables the system to identify potential error sources in the communication link based on multimodal fusion, improving the spatial coverage and temporal synchronization of data acquisition. At the same time, it outputs various raw signals to the data acquisition layer through the CAN bus, providing high-quality data support for subsequent error compensation and digital twin model construction, thereby significantly enhancing the adaptability, error suppression capability, and status visualization level of the communication system in complex environments, and realizing high-reliability perception of communication nodes and intelligent monitoring of connection structures.

[0054] It should be further explained that when the communication equipment is in a static calibration state, the data acquisition layer acquires point cloud and image data of the communication nodes of the static structural components. It uses a calibration template and ICP algorithm to uniformly register the data acquired from different angles to the global coordinate system, constructs a static reference model, and stores it in the database. During the operation and testing of the communication equipment, the data acquisition layer adopts an exposure mechanism with an exposure time of less than 5ms, combined with synchronously triggered pulse lighting, to instantly freeze the dynamic process. The dynamically acquired data is divided into several slices, which are processed in parallel by each worker node in the Spark platform. At the same time, environmental disturbance data, including temperature, humidity, vibration, and illumination parameters, are collected synchronously to construct a time series for error modeling. By aligning the dynamic data with the static reference model, the dynamic error characteristics of the communication node caused by vibration, thermal expansion, and structural displacement during operation are extracted.

[0055] By acquiring point cloud and image data of communication nodes under static calibration conditions, and using calibration templates and ICP algorithms to uniformly register multi-angle data to a global coordinate system to construct a static reference model, a high-speed exposure mechanism with an exposure time of less than 5ms and pulsed illumination are combined during equipment operation testing to achieve instantaneous freezing of the dynamic process. The dynamically acquired data is then divided into slices and processed in parallel by worker nodes on the Spark platform. Simultaneously, disturbance parameters such as temperature, humidity, vibration, and illumination are collected to construct a time series of environmental changes. This allows the system to align the dynamically acquired data with the static model with high precision, thereby accurately extracting the dynamic error changes of communication nodes caused by factors such as vibration disturbance, thermal expansion, and structural displacement during operation. This enables error spatiotemporal distribution modeling and correlation analysis, significantly improving the system's sensitivity to changes in operating status and its error compensation capability. This provides data support for subsequent image correction, spatial coordinate updates, and digital twin updates, enhancing the stability, accuracy, and intelligent analysis capabilities of the communication system under dynamic operating conditions.

[0056] It should be further explained that, in the data acquisition layer, the static structural components acquired include, but are not limited to, fiber optic patch cord interfaces, base station transceiver connectors, feeder channel ports, antenna power supply devices, and key connection points of RF amplifier modules.

[0057] In communication equipment structures, components such as fiber optic patch cord interfaces, base station transceiver connectors, feeder channel ports, antenna feeding devices, and RF amplifier modules are typical high-frequency, state-sensitive, and critical connection nodes that significantly impact communication stability. During long-term operation, these components are highly susceptible to structural deformation, loosening of contacts, or connection misalignment due to mechanical stress, thermal strain, or vibration shock, directly affecting signal integrity and transmission quality. Therefore, this system focuses on sensing the spatial state of these structural components at the data acquisition layer, which helps in building a benchmark model and monitoring their dynamic errors. However, considering that communication systems may integrate more types of connection units (such as power dividers, couplers, photoelectric conversion modules, MIMO array components, etc.) in different scenarios, limiting the system to only the aforementioned components would restrict its adaptability and technical scalability. Therefore, this invention uses the description "including but not limited to," ensuring that the system prioritizes the aforementioned typical connection points while allowing subsequent expansion to other types of communication structural components. This ensures that the technical solution has broader adaptability and engineering scalability in actual deployment, while avoiding unnecessary limitations.

[0058] It should be noted that the software system layer uses the raw data from the data acquisition layer to calculate the environmental impact factor according to the preset environmental compensation model, corrects the image preprocessing algorithm and optimizes the filter covariance parameter, dynamically adjusts the correction strategy according to the disturbance values ​​of temperature, humidity and vibration, calculates the communication structure error based on the deviation between the static model and the real-time data and smooths the error trajectory through Kalman filtering, updates the node coordinates in real time, and trains the autoencoder network on the Spark platform through stochastic gradient descent to perform feature extraction and noise suppression.

[0059] By utilizing raw data acquired from the data acquisition layer at the software system layer and dynamically calculating environmental impact factors using a pre-defined environmental compensation model, and adaptively adjusting the image preprocessing algorithm and filter covariance parameters, the system can correct the error information of communication nodes in real time based on disturbance values ​​such as temperature, humidity, and vibration. This effectively offsets the impact of external disturbances on image quality, point cloud alignment, and node positioning accuracy. Simultaneously, based on the deviation between the static reference model and dynamic real-time data, the system calculates communication structure errors and uses Kalman filtering to smooth and predict error trajectories, achieving dynamic high-precision updates of node coordinates. Furthermore, the system deploys an autoencoder training process on the Spark platform, combining stochastic gradient descent algorithms for feature extraction and high-dimensional noise reduction of the acquired data, further improving data quality and system robustness. Overall, this strategy implements a multi-layered, interconnected compensation mechanism for environmental disturbance modeling, error estimation, and noise reduction learning, significantly enhancing the data communication system's adaptability, positioning stability, and noise tolerance in complex dynamic environments.

[0060] It should be noted that in the data acquisition layer, the environmental compensation model is based on simulated communication scene data. A linear weighted model is constructed through regression analysis, and image enhancement parameters are corrected according to illumination, temperature, humidity, and vibration disturbances. The exposure time and gain are calculated as follows:

[0061]

[0062] Where T and G are the exposure value and gain value, respectively, T0 and G0 are the original exposure value and original gain value, respectively, and W... L As a perturbation weighting factor, the environmental monitoring module collects current illumination, current ambient temperature, current relative humidity, and current vibration acceleration in real time during system operation. During the offline calibration phase of the system, the mapping relationship between several sets of environmental variables and the optimal image exposure value is collected to form a sample set. A weighting model is established for the perturbation variables using the multiple linear regression method. The weighted sum of the differences between the current value and the standard value of each perturbation variable is used as the perturbation weighting factor. The weight coefficients of the differences between the current value and the calibration value of each perturbation variable are obtained through training.

[0063] By constructing an environmental compensation model based on simulated communication scenario data in the data acquisition layer, the system uses regression analysis to establish the response mapping relationship between disturbance parameters such as illumination, temperature, humidity, and vibration and the optimal exposure value of the image. Furthermore, it employs a multiple linear regression method to extract the influence weights of various disturbances on image acquisition accuracy, constructing a disturbance weighting factor W. L This enables dynamic correction of exposure time T and gain G, allowing the system to adjust image acquisition settings in real time to achieve optimal contrast and clarity even under conditions of drastic changes in on-site environmental parameters. It effectively avoids image blurring and recognition errors caused by light fluctuations, thermal expansion and contraction, humidity changes, or equipment vibration, improves data acquisition stability and consistency across time periods, enhances the adaptability and robustness of image processing to on-site interference, and provides higher-quality basic data support for communication component structure recognition, attitude tracking, and subsequent digital twin construction.

[0064] Example

[0065] To clearly illustrate the calculation method of exposure time and gain in the data acquisition layer, the following examples will be used for detailed explanation.

[0066] To clearly illustrate the exposure time and gain calculation method at the data acquisition layer, and in conjunction with the communication system operation flow of this invention, the following embodiments will provide a detailed explanation:

[0067] In an industrial communication test platform, a multimodal sensing terminal with image acquisition capabilities is deployed. This terminal is used to monitor the status and structural dynamic changes of the radio frequency interface of the base station transceiver module. The on-site environment exhibits significant disturbance characteristics, such as high-frequency vibrations (0.8–1.5 m / s²). 2 Temperature fluctuations (15–35℃), humidity drift (40–80% RH), and light changes (200–1500 lux).

[0068] The system first collects several sets of standard communication component image data and corresponding environmental disturbance variables during the offline calibration phase to construct a training sample set (L). i ,K i H i A i ), where i is the index of the sample set, L i T i K i A i These are the i-th high-frequency vibration training sample, the i-th temperature fluctuation training sample, the i-th humidity drift training sample, and the i-th light change training sample, respectively. The optimal value of the training sample is T. opt,i G opt,i T opt,i G opt,iThe optimal exposure and optimal gain values ​​were collected under this perturbation condition, respectively. The perturbation weight coefficients were trained using a multiple linear regression method, and the fitted model is as follows:

[0069] W L =ω1(L-L0)+ω2(K-K0)+ω3(H-H0)·+ω4(A-A0)

[0070] Where ω1, ω2, ω3, and ω4 are the weighting coefficients for each disturbance, L, K, H, and A are the light intensity, temperature, humidity, and vibration acceleration, respectively, and L0, K0, h0, and A0 are the baseline values ​​for L, K, H, and A, respectively. During the real-time operation of the system, the environmental monitoring module collects the current disturbance parameters and dynamically calculates the weighting factor W. L And substitute it into the image parameter adjustment formula:

[0071]

[0072] Where T0 and G0 are the original exposure value and the original gain value, respectively, and T0 = 8ms and G0 = 1.2 are the initial exposure parameters and gain parameters under static calibration conditions.

[0073] After on-site testing and verification in four groups, the data obtained are shown in Table 1 below:

[0074] Table 1. Four sets of on-site test data

[0075]

[0076] In Table 1, the unit for light intensity is lux, the unit for temperature is ℃, the unit for humidity is relative humidity, and the unit for vibration is m / s. 2 W L The values ​​are obtained by weighting the perturbation variable with its deviation value, with weights ω1, ω2, ω3, and ω4 of 0.0005, 0.02, 0.015, and 0.1, respectively. The reduction in blur rate and the improvement in contrast are evaluated by an image sharpness rating function (here, a weighted average of the gradient function and the contrast entropy function is used, with weights of 0.5 and 0.5, respectively). As can be seen from the data:

[0077] In environments with strong disturbances (such as Group 3), the correction mechanism adjusts the exposure time from 8ms to 4.97ms, significantly suppressing overexposure and blur.

[0078] In scenes with weak disturbances or low light (such as Group 2), the system automatically extends the exposure and increases the gain to ensure the integrity of the image signal.

[0079] The image blur rate decreased by more than 18% in all scenarios, and the contrast was significantly improved, which verified the environmental adaptability and robustness of the model.

[0080] It should be noted that at the software system level, on the Spark platform, the autoencoder training process includes:

[0081] Step 1: The master node broadcasts the network size, random number seed τ, and learning rate η to each worker node, and at the same time divides the preprocessed data into shards;

[0082] Step 2: Each Worker independently builds an autoencoder to perform forward propagation, error backpropagation, and parameter updates on the samples;

[0083] Step 3: Each worker uploads its local parameters to the master node, and the master node aggregates, updates, and broadcasts the new model;

[0084] Step 4: The trained model generates a correction factor to correct the coordinates of the communication node data.

[0085] It should be further explained that in step 1, each Worker node synchronously receives multi-view image streams and IMU and optical flow data provided by the auxiliary imaging unit to construct a multimodal dataset;

[0086] In step 2, the spatiotemporal consistency mask of the image is calculated by combining optical flow and IMU, occlusion features are predicted and missing regions are compensated, and the fused feature data is input into the autoencoder for training.

[0087] By introducing a distributed computing structure into the autoencoder training process deployed on the Spark platform, the master node and each worker node can collaborate efficiently to achieve parallel learning and iterative optimization of multi-source communication data. In step 1, each worker node synchronously receives multi-view image streams, IMU data, and optical flow information collected by the auxiliary imaging unit, constructing a multimodal dataset that integrates image temporal features and spatial pose information, effectively enhancing the model's ability to perceive communication structure occlusion, viewpoint shift, and local information loss. In step 2, a spatiotemporal consistency mask between images is calculated by combining IMU and optical flow data, which can identify key feature loss regions caused by motion or occlusion, and perform inter-frame compensation interpolation to ensure that the input data maintains structural continuity and spatial consistency. Finally, the compensated multimodal features are fed into the autoencoder training network for deep feature extraction and noise suppression, which not only improves the model's robustness and fault tolerance to abnormal data, but also significantly improves the spatial coordinate stability and data accuracy of communication nodes in dynamic environments, providing highly reliable and disturbance-resistant deep representation features for subsequent digital twin mapping and error compensation.

[0088] It should be noted that the functional application layer uses the corrected data to construct a three-dimensional model of the communication system, and displays the status changes, error trends and key indicator curves of communication nodes in real time through the digital twin platform, supporting alarm output, operation traceability and automatic report generation.

[0089] By integrating and modeling the corrected multi-source data through the functional application layer, a three-dimensional structural model of the communication system is constructed. This model is then connected to a digital twin platform to map the spatial status, error trends, and key communication parameter change curves of the communication nodes in real time. This achieves high-precision dynamic synchronization from physical entities to digital models, enabling the system to visualize the operational status of the communication structure, capture abnormal fluctuations in real time, and analyze trends. It supports triggering alarms by node dimension, tracing historical operating condition changes, and automatically generating structural offset reports and environmental interference analysis results. This effectively improves the operational transparency, maintenance efficiency, and fault response capabilities of the communication system, meeting the needs for high-precision status monitoring and intelligent decision support in complex industrial communication scenarios.

[0090] In summary, this invention, by introducing multimodal perception, dynamic error compensation, environmental disturbance modeling, and autoencoder training mechanisms, achieves a high degree of integration of data acquisition, synchronization, modeling, and prediction for communication nodes in complex industrial environments. It can identify in real time error fluctuations caused by disturbances such as temperature, humidity, vibration, and illumination, as well as micro-offsets in the communication structure. A dynamic error field is constructed based on the alignment of static models and dynamic data. Combined with Kalman filtering and linear regression to dynamically correct image exposure and gain parameters, it effectively improves the clarity, stability, and transmission accuracy of communication images and point clouds. Simultaneously, the autoencoder is trained in parallel using the Spark platform to complete high-dimensional noise suppression and feature extraction, giving the system stronger anti-interference capabilities, adaptive adjustment capabilities, and online modeling capabilities. Finally, a digital twin communication model is constructed through the functional application layer, achieving high-fidelity data transmission and intelligent operation and maintenance in industrial-grade communication scenarios while ensuring data synchronization, link stability, and fault prediction.

[0091] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0092] In conclusion, the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data communication system based on multi-source data synchronization and dynamic error compensation, comprising a hardware device layer, the hardware device layer including a multimodal sensing terminal, an auxiliary imaging acquisition unit, a laser scanning device, and an environmental monitoring module, characterized in that, The multimodal sensing terminal, auxiliary imaging acquisition unit, laser scanning device, and environmental monitoring module are all connected to data acquisition cards. The hardware device layer is connected to the data acquisition layer, the data acquisition layer is connected to the software system layer, and the software system layer is connected to the functional application layer. The hardware device layer is used to collect structural point cloud data and environmental information of communication equipment components in static and dynamic states. The data acquisition layer extracts errors in the dynamic operation of the communication equipment by analyzing the static point cloud information. The software system layer uses the static benchmark model and environmental monitoring data to dynamically correct and compensate for the errors, and corrects the positioning coordinates of the communication nodes through multi-source fusion and filtering technology. The functional application layer constructs a three-dimensional model of the communication components and a digital twin system based on the corrected data to realize remote monitoring, communication flow analysis, and anomaly early warning.

2. The data communication system based on multi-source data synchronization and dynamic error compensation according to claim 1, characterized in that, The multimodal sensing terminal includes a sensing head component with a reflective target mounted on it. An imaging lens is mounted on one side of the reflective target, connected to an image sensor. The image sensor is connected to a signal processing module. The surface of the reflective target has an anti-interference multilayer film structure, which, from the inside out, includes: First coating: a silicon nitride interference film with a thickness of 85 nm; Second coating: an aluminum oxide electromagnetic shielding layer with a thickness of 50nm; Third coating: a silicon carbide structural film with a thickness of 90nm; Fourth coating: a 55nm thick zinc oxide antistatic film; Fifth coating: Magnesium fluoride antireflective film with a thickness of 85nm.

3. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 2, characterized in that, In the hardware device layer, the sensor head component senses the connection status, spatial changes, and relative posture information of the communication interface area. The auxiliary imaging acquisition unit captures the reflected mark image for visual reference positioning from a set angle. The laser scanning device supplements the acquisition of the three-dimensional point cloud information of the communication component and identifies the connection status and spatial configuration of the communication component. The environmental monitoring module collects disturbance data of temperature, humidity, and vibration for use as a dynamic compensation reference. The data acquisition card collects the raw signal and outputs it to the data acquisition layer through the CAN bus. The communication node structure information includes, but is not limited to, the edge contour, contact curvature, and transition area morphology of the fiber optic interface, RF connector, and cable terminal. The image acquisition direction is based on the normal direction of a communication port, and the viewing angle is set every 2° within a range of ±30°.

4. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 1, characterized in that, When the communication equipment is in a static calibration state, the data acquisition layer acquires point cloud and image data of the communication nodes of the static structural components. It uses a calibration template and ICP algorithm to uniformly register the data acquired from different angles to the global coordinate system, constructs a static reference model, and stores it in the database. During the operation and testing of the communication equipment, the data acquisition layer adopts an exposure mechanism with an exposure time of less than 5ms, combined with synchronously triggered pulsed illumination, to instantly freeze the dynamic process. The dynamically acquired data is divided into several slices, which are processed in parallel by each worker node in the Spark platform. At the same time, environmental disturbance data, including temperature, humidity, vibration, and illumination parameters, are collected synchronously to construct a time series for error modeling. By aligning the dynamic data with the static reference model, the dynamic error characteristics of the communication nodes caused by vibration, thermal expansion, and structural displacement during operation are extracted.

5. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 4, characterized in that, In the data acquisition layer, the static structural components acquired include, but are not limited to, fiber optic patch cord interfaces, base station transceiver connectors, feeder channel ports, antenna power supply devices, and key connection points of RF amplifier modules.

6. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 2, characterized in that, The software system layer utilizes the raw data from the data acquisition layer to calculate environmental impact factors based on a preset environmental compensation model. It corrects the image preprocessing algorithm and optimizes the filter covariance parameters. It dynamically adjusts the correction strategy based on the disturbance values ​​of temperature, humidity, and vibration. Based on the deviation between the static model and real-time data, it calculates the communication structure error and smooths the error trajectory through Kalman filtering. It updates the node coordinates in real time and trains an autoencoder network on the Spark platform through stochastic gradient descent for feature extraction and noise suppression.

7. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 1, characterized in that, In the data acquisition layer, the environmental compensation model is based on simulated communication scene data. A linear weighted model is constructed through regression analysis, and image enhancement parameters are corrected according to illumination, temperature, humidity, and vibration disturbances. The exposure time and gain are calculated as follows: Where T and G are the exposure value and gain value, respectively, T0 and G0 are the original exposure value and original gain value, respectively, and W... L As a perturbation weighting factor, the environmental monitoring module collects current illumination, current ambient temperature, current relative humidity, and current vibration acceleration in real time during system operation. During the offline calibration phase of the system, the mapping relationship between several sets of environmental variables and the optimal image exposure value is collected to form a sample set. A weighting model is established for the perturbation variables using the multiple linear regression method. The weighted sum of the differences between the current value and the standard value of each perturbation variable is used as the perturbation weighting factor. The weight coefficients of the differences between the current value and the calibration value of each perturbation variable are obtained through training.

8. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 7, characterized in that, At the software system level, on the Spark platform, the autoencoder training process includes: Step 1: The master node broadcasts the network size, random number seed τ, and learning rate η to each worker node, and at the same time divides the preprocessed data into shards; Step 2: Each Worker independently builds an autoencoder to perform forward propagation, error backpropagation, and parameter updates on the samples; Step 3: Each worker uploads its local parameters to the master node, and the master node aggregates, updates, and broadcasts the new model; Step 4: The trained model generates a correction factor to correct the coordinates of the communication node data.

9. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 8, characterized in that, In step 1, each Worker node synchronously receives multi-view image streams and IMU and optical flow data provided by the auxiliary imaging unit to construct a multimodal dataset; In step 2, the spatiotemporal consistency mask of the image is calculated by combining optical flow and IMU, occlusion features are predicted and missing regions are compensated, and the fused feature data is input into the autoencoder for training.

10. A data communication system based on multi-source data synchronization and dynamic error compensation according to claim 1, characterized in that, The functional application layer uses the corrected data to construct a three-dimensional model of the communication system. Through the digital twin platform, it displays the status changes, error trends and key indicator curves of communication nodes in real time, and supports alarm output, operation traceability and automatic report generation.

Images